Varun Nair

I wrote about how I’m using our own product to move faster and to quickly distill what I need to focus on.

Context is a bottleneck when working with LLMs, not prompt engineering.

My latest weekly post: garbage in, garbage out. Dumping data into a LLM's context is like recording in a noisy room – you end up with competing signals that steer the output into vague generalities. So, how do you maximise the signal to noise ratio?

If it takes the Olympics to build ambitious public transport projects, so be it! Kudos to the French for pulling off a great Olympics.

Deleting code feels good but what's even better? Ditching infrastructure. For over a year, we used high-level cloud services to move fast. They were easy to set up but soon the complexity piled on: multiple services, abstractions, and management overhead. It reached a tipping point where adding more services wasn't the answer – it was time to shed the weight. Farewell, app services, functions, service buses, and the like. Welcome back, Docker and Kubernetes.

And here’s a video interview of him (from 1929!) youtu.be/o2okclRid4M?si=uvl_…

Ah yes, I had forgotten how slow banking in America is.

Why does Teams require more than 1gb of disk space?!

Meta’s stock market comeback this year has been pretty phenomenal